24 MPx + 18-105 = 19 MPx

Started 10 months ago | Discussions thread
Flat view
TJ61 Senior Member • Posts: 1,320
24 MPx + 18-105 = 19 MPx

Hey, that's just math!

Earlier, I reported how I'd either stumbled across Sony's dirty little secret, or finally learned what everyone else knows about distortion correction. Until then, my thinking on lens distortion had been, if it can be corrected, what do I care?

To further investigate the real effect of distortion correction, I determined how much cropping goes on during correction, and plotted the following Corrected Resolution as a function of focal length for the E PZ 18-105 on my a6000:

Reduced resolution due to rectangular cropping after distortion correction

To make this plot, I took a number of exposures with my 18-105 on my a6000, adjusting the focal length from 18 to 105mm. I imported the raw photos into Capture One, and applied their lens profile (and not the manufacturer's) to correct distortion. The correction process first applies a transform that appears to pull in the four corners towards the center. Then it crops the now-bulging edges back to rectangular. The plotted data is the number of pixels remaining after this cropping.

The manufacturer's profile (supplied by Sony, as opposed to the one made by the folks at C1) works largely the same way. But, just like the in-camera jpg processor, I guess it uprezzes the image back to 6000x4000, and you're none the wiser. Being of the opinion that any such processing could only add noise to the data, I now prefer the C1 lens profile, which stops after the cropping. I certainly don't want to do any uprezzing before I do my other PP. So, for that reason I'm glad I figured this out.

I also think a plot like this is the clearest way to show the primary effect of distortion in a lens; a quantifiable pixel reduction due to distortion correction. While it's not surprising that distortion correction might require throwing out some pixels, I was a little surprised that it's as much as 20% for this lens, and it drops to that value quite fast.

Now, for comparison, here's the same analysis applied to my other lenses:

Apparent resolution change due to rectangular cropping after distortion correction for various lenses

Note that a couple lenses actually appear to give increased resolution when corrected for distortion. Of course, that's not "real". But I'd rather see uprezzing beyond 24 MPx (and subsequent downrezzing back to 24 MPx), than an outright loss of pixels (even though the former must add a little mathematical noise).  Otherwise, the 18-105 stands alone in terms of amount of resolution lost.

I want to be very clear that this is not meant as a lens-bashing post. I'm as fond as ever of my 18-105; the workhorse of my small stable of lenses. I think what I'm showing here might be the most straight-forward way to show the primary effect of such distortion in this lens -- the quantified loss of pixels due to distortion correction.

 TJ61's gear list:TJ61's gear list
Sony a6000 Sony E 10-18mm F4 OSS Sony E 35mm F1.8 OSS Sony E PZ 18-105mm F4 G OSS
Sony a6000
If you believe there are incorrect tags, please send us this post using our feedback form.
Flat view
Keyboard shortcuts:
FForum PPrevious NNext WNext unread UUpvote SSubscribe RReply QQuote BBookmark MMy threads
Color scheme? Blue / Yellow